Importance of good speech reproduction

Discussion in 'Hi-Fi and General Audio' started by RobHolt, Mar 29, 2011.

  1. RobHolt

    RobHolt Moderator

    Joined:
    Oct 24, 2004
    Messages:
    4,126
    Likes Received:
    9
    Decades ago, men in white coats at the BBC spent much of their time designing speaker systems with the accurate reproduction of human voice as the ultimate goal.
    How well they succeeded in their mission is down to the individual listener (I think they got it largely right) but having heard a great many loudspeakers, some really do destroy the tonality and coherence required to get speech right.

    A good clean voice recording can really highlight cone/dome/cabinet related resonances and will also highlight any crossover problems. It can highlight the various honks and quacks that make loudspeaker reproduced voice unnatural.

    On the flip side, some emphasis around the presence range, while detracting from ultimate sonic fidelity can actually improve intelligibility to some listeners.

    Is this aspect of performance important to you?

    To help you assess the performance of your loudspeakers, here is a short speech recording made by Alan Shaw and Derek Hughes (Harbeth) at the old BBC Kingswood Warren anechoic chamber. This is a good clean mono speech recording and a stern test for your system. I can't see any restriction on use and it is freely available to download.

    Have a listen - the discussion is also very interesting, as is the recording venue as this was used to develop many of the famous BBC speaker designs.

    http://www.audiosmile.com/forum/soundfiles/Speech.mp3
     
    RobHolt, Mar 29, 2011
    #1
  2. RobHolt

    Tenson Moderator

    Joined:
    Nov 12, 2003
    Messages:
    5,947
    Likes Received:
    1
    Location:
    Kent, UK
    Do you think this should be played stereo or mono?

    Mono will surely sound more real but I don't think it is such a good test of your hi-fi since voices are usually stereo in centre stage. The tonality element of a speaker might be great and sound good in mono, but if the speaker pair doesn't produce a good soundstage in your room the illusion could be broken.

    For me, a good centre image is probably just as important to making a recoding of voice sound real as tonality is.
     
    Tenson, Mar 29, 2011
    #2
  3. RobHolt

    RobHolt Moderator

    Joined:
    Oct 24, 2004
    Messages:
    4,126
    Likes Received:
    9
    Mono definitely, though it really is intended to test for colouration.
     
    RobHolt, Mar 30, 2011
    #3
  4. RobHolt

    Alan Brown

    Joined:
    Jul 2, 2008
    Messages:
    78
    Likes Received:
    0
    I've never really 'got on' with BBC inspired speakers, though far be it from me to knock such a highly developed discipline and engineering quality. I actually live 100 yards from Spendor and have often heard their designs, I also am quite familiar with B&W gear and my first proper HiFi speakers were KEF Cadenzas.

    I guess I am more in the 'RD' school of thought on speakers which dictates 'less is more'. I love my Royd RR3s which are sublimely engineered to be simple, if you know what I mean. I also have some Goodmans Magisters in my garage, which I must get out one day, as well as a 'spare' set of Minstrels, and some AR18 in the loft. So I find that for me, a speaker with a simpler crossover just plays music better, and certainly seems to let amplifiers to their thing better (though my current amp is - predictably - picky about this sort of thing).

    The thing is, all these 'simple' speakers reproduce voice really well, especially when recorded on CD. I have found on some occasions that BBC digital TV speech is less than perfect though - this may be down to the SKY box perhaps, or very low bit rates from a cost cutting BBC news dept (the 'red button' weather forecast looks & sounds like it was recorded on a camera phone - 'This Week' on Thu night is similarly poor).

    I don't know if human speech is the 'correct' way to ascertain a speakers competence, reproducing this is very different to playing music. To say - as some have - that 'if a speaker can reproduce voice really well it can play everything' seems to be a rather large leap to make. I am quite prepared to compromise on speech if the musical abilities of the speaker are increased, though I don't know how to quantify either of those.

    The thing is I have always found my speakers can do the job perfectly, without ever assessing them in this area seriously. I will be quite interested to download the Harbeth sample & see what it tells me.
     
    Alan Brown, Mar 31, 2011
    #4
  5. RobHolt

    Janko

    Joined:
    Mar 28, 2011
    Messages:
    41
    Likes Received:
    0
    From my experience pasive speakers cannot compete with active in terms of colouration, distortion.... (at least in the same price range)
    There are also many companies who intentionally colour the sound of hifi speakers to make it sound more "attractive"... which rarely happens in pro audio.
    Sadly great majority of hifi buffs is on passive speakers. It's probably more a matter of taste and not quality of sound reproduction.
     
    Janko, Mar 31, 2011
    #5
  6. RobHolt

    UK Duty Paid

    Joined:
    Aug 17, 2009
    Messages:
    155
    Likes Received:
    0
    Thanks I'll download & listen

    I put great store in speech & find when this feels right everything else falls into place.
    Some systems & tuners seem to inflict an electronic element into speech so you are aware its a recording. Obviously doesn't happen in reality.
     
    UK Duty Paid, Mar 31, 2011
    #6
  7. RobHolt

    Coda II getting there slowly

    Joined:
    Nov 17, 2004
    Messages:
    603
    Likes Received:
    1
    Location:
    Devon
    Have you had a look at the attachment to post #5 of this thread on HUG?

    It's a lexicon of loudspeaker excesses and deficiencies and how they translate into subjective terms.

    I was struck by the comments below 300Hz, eg "Good speech, thin unnatural music".
     
    Coda II, Apr 1, 2011
    #7
  8. RobHolt

    Tenson Moderator

    Joined:
    Nov 12, 2003
    Messages:
    5,947
    Likes Received:
    1
    Location:
    Kent, UK
    I think it is somewhat difficult to assess true accuracy with an unfamiliar voice, but it does tend to highlight obvious tonal colourations like a rise in the presence range or poor crossover integration.

    Rob and I once did a test with the Kensai like this. I made a recording of myself in anechoic conditions reading from a magazine, then I sat next to the Kensai in Robs living room and read the same piece. Rob was then able to compare the recording of me via Kensai, to the real me. The result was surprisingly similar sound. The tone was very natural with no noticeable colouration.

    However, the perception of the sound was slightly different, and I'd put this down to the dispersion of the speaker vs. my real voice. The speaker might have a perfect frequency response and zero distortion, but it will radiate sound in a different shape than a humans body, throat, mouth combination. This makes the perception of the sound different in a room, even though tone and distortion might be ideal.

    The only way to over-come this is to replay the loudspeakers and the real voice in an anechoic chamber, but maybe that is somewhat missing the point since it won't be used like that.

    I think the point here is really that instruments have different dispersion patterns than a speaker. I read some marketing blurb about a panel speaker recently that said it was developed originally to work in an electronic piano with dispersion to match a real piano. They tried to then make the link that it would sound more like real instruments because of this, but of course every instrument has a different dispersion so the theory doesn't hold.

    You also have to wonder if it's better to have a singer sound like they are in the room with you, or move you to their room. The former is easier for our ear/mind to identify with as every other sense backs up that we are in this room. I feel the later is the real goal though, and as listeners we need to learn how to listen only with our ears and let them take us to the recording venue, despite contrary information from our sight, smell sense, touch etc.. Speakers are capable of the this, but real accurate reproduction of the former requires matching dispersion patterns of the original source.

    I think, therefore this speech test is floored, as it is trying to make a speaker sound like a voice in your room, rather than transport the listener to the room where the voice was originally.

    I hope that all makes some sense, I'm just typing as I go!
     
    Tenson, Apr 1, 2011
    #8
  9. RobHolt

    RobHolt Moderator

    Joined:
    Oct 24, 2004
    Messages:
    4,126
    Likes Received:
    9
    We did the same thing with the old Quad ESL - and that was superb on voice but only on a very narrow axis.

    This test is, IMO, useful for assessing basic accuracy and identifying colouration, and very good for highlighting crossover issues.

    The Harbeth recording sounds very obviously wrong on some speakers. Exposure to panel (and in this case electrostatic) speakers does i think heighten sensitivity to vocal colourations.
     
    RobHolt, Apr 1, 2011
    #9
  10. RobHolt

    Janko

    Joined:
    Mar 28, 2011
    Messages:
    41
    Likes Received:
    0
    Isn't the former what omnidirectional speakers can do. It is still not the real thing, a bit closer to a ghost in your room but in your room nevertheless. Well an omnidirectional tweeter is enough to produce that effect, since lower frequencies are more omnidirectional anyway.
    Janko
     
    Janko, Apr 2, 2011
    #10
Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.